🤖 AI Summary
This study addresses the fundamental tension between user utility and bystander privacy in smart glasses, particularly in sensitive contexts where privacy expectations and behavioral intentions diverge significantly. Through a large-scale survey (N=525) and paired interviews (N=20), it systematically evaluates multi-stakeholder acceptance of twelve privacy-enhancing technologies, uncovering four structural privacy trade-offs. Findings reveal that 65%–90% of bystanders exhibit defensive behaviors in sensitive scenarios, confirming context as a critical determinant of privacy acceptability. Building on these insights, the work proposes the first context-driven adaptive privacy framework for smart glasses, moving beyond static design paradigms. It further introduces a context-aware diagnostic model for privacy mechanisms, enabling three dynamic, situation-responsive privacy protection strategies tailored to real-world deployment.
📝 Abstract
Camera glasses create fundamental privacy tensions between wearers seeking recording functionality and bystanders concerned about unauthorized surveillance. We present a systematic multi-stakeholder evaluation of privacy mechanisms through surveys (N=525) and paired interviews (N=20) in China. Study 1 quantifies expectation-willingness gaps: bystanders consistently demand stronger information transparency and protective measures than wearers will provide, with disparities intensifying in sensitive contexts where 65-90% of bystanders would take defensive action. Study 2 evaluates twelve privacy-enhancing technologies, revealing four fundamental trade-offs that undermine current approaches: visibility versus disruption, empowerment versus burden, protection versus agency, and accountability versus exposure. These gaps reflect structural incompatibilities rather than inadequate goodwill, with context emerging as the primary determinant of privacy acceptability. We propose context-adaptive pathways that dynamically adjust protection strategies: minimal-friction visibility in public spaces, structured negotiation in semi-public environments, and automatic protection in sensitive contexts. Our findings contribute a diagnostic framework for evaluating privacy mechanisms and implications for context-aware design in ubiquitous sensing.